In [1]:
import numpy as np
import pandas as pd
import math
import cv2
from matplotlib import pyplot as plt
import matplotlib.cm as cm
import matplotlib.gridspec as gridspec
from pylab import imread
from skimage.color import rgb2gray
In [2]:
def ShowImage(ImageList, nRows = 1, nCols = 2, WidthSpace = 0.00, HeightSpace = 0.00):
from matplotlib import pyplot as plt
import matplotlib.gridspec as gridspec
gs = gridspec.GridSpec(nRows, nCols)
gs.update(wspace=WidthSpace, hspace=HeightSpace) # set the spacing between axes.
plt.figure(figsize=(20,10))
for i in range(len(ImageList)):
ax1 = plt.subplot(gs[i])
ax1.set_xticklabels([])
ax1.set_yticklabels([])
ax1.set_aspect('equal')
plt.subplot(nRows, nCols,i+1)
image = ImageList[i].copy()
if (len(image.shape) < 3):
plt.imshow(image, plt.cm.gray)
else:
plt.imshow(image)
plt.title("Image " + str(i))
plt.axis('off')
plt.show()
In [3]:
def InfoImage(IM):
Image = IM
Width = Image.shape[1]
Height = Image.shape[0]
Channel = len(Image.shape)
print("Width : ", Width, " Height : ", Height, " Channel : ", Channel)
if(Channel == 2):
print("Min Intensity: ", IM.min(), " Max Intensity: ", IM.max())
else:
print("Red - Min Intensity: ", IM[:,:,0].min(), " Max Intensity: ", IM[:,:,0].max())
print("Green - Min Intensity: ", IM[:,:,1].min(), " Max Intensity: ", IM[:,:,1].max())
print("Blue - Min Intensity: ", IM[:,:,2].min(), " Max Intensity: ", IM[:,:,2].max())
In [4]:
def ResizeImage(IM, DesiredWidth, DesiredHeight):
from skimage.transform import rescale, resize
OrigWidth = float(IM.shape[1])
OrigHeight = float(IM.shape[0])
Width = DesiredWidth
Height = DesiredHeight
if((Width == 0) & (Height == 0)):
return IM
if(Width == 0):
Width = int((OrigWidth * Height)/OrigHeight)
if(Height == 0):
Height = int((OrigHeight * Width)/OrigWidth)
dim = (Width, Height)
# print(dim)
resizedIM = cv2.resize(IM, dim, interpolation = cv2.INTER_NEAREST)
# imshows([IM, resizedIM], ["Image", "resizedIM"],1,2)
return resizedIM
1/ Kiếm 1 ảnh từ internet thực hiện các thao tác biến đổi sau bằng cách gọi hàm từ các thư viện (opencv hay skimage hay bất kỳ thư viện nào)¶
- Scaling
- Rotation (30, 45, 60, 90, 120, 160, 180)
- Translation
- Shearing
In [5]:
# Read Image
image_color = imread("Sample08/giraffe.jpg")
# Convert Image into Gray
image_gray = cv2.cvtColor(image_color, cv2.COLOR_RGB2GRAY)
# Display Image
ShowImage([image_color, image_gray], 1, 2)
In [6]:
# Scaling
image_scale01 = cv2.resize(image_color,None,fx=0.5, fy=1, interpolation = cv2.INTER_CUBIC)
image_scale02 = cv2.resize(image_color,None,fx=0.5, fy=0.5, interpolation = cv2.INTER_CUBIC)
image_scale03 = cv2.resize(image_color,None,fx=1, fy=0.5, interpolation = cv2.INTER_CUBIC)
print("\n image_color :")
InfoImage(image_color)
print("\n image_scale01 :")
InfoImage(image_scale01)
print("\n image_scale02 :")
InfoImage(image_scale02)
print("\n image_scale03 :")
InfoImage(image_scale03)
ShowImage([image_scale01, image_scale02, image_scale03], 1, 3)
In [7]:
# Translation
rows,cols = image_color.shape[:2]
TranslationMatrix = np.float32([[1,0,200],[0,1,0]])
image_translation01 = cv2.warpAffine(image_color,TranslationMatrix,(cols,rows))
TranslationMatrix = np.float32([[1,0,0],[0,1,200]])
image_translation02 = cv2.warpAffine(image_color,TranslationMatrix,(cols,rows))
TranslationMatrix = np.float32([[1,0,200],[0,1,200]])
image_translation03 = cv2.warpAffine(image_color,TranslationMatrix,(cols,rows))
TranslationMatrix = np.float32([[1,0,-200],[0,1,-200]])
image_translation04 = cv2.warpAffine(image_color,TranslationMatrix,(cols,rows))
ShowImage([image_translation01, image_translation02], 1, 2)
ShowImage([image_translation03, image_translation04], 1, 2)
In [8]:
# Rotation
rows,cols = image_color.shape[:2]
RotationMatrix = cv2.getRotationMatrix2D((cols/2,rows/2),90,1)
image_rotation01 = cv2.warpAffine(image_color,RotationMatrix,(cols,rows))
RotationMatrix = cv2.getRotationMatrix2D((cols/2,rows/2),45,1)
image_rotation02 = cv2.warpAffine(image_color,RotationMatrix,(cols,rows))
RotationMatrix = cv2.getRotationMatrix2D((cols/2,rows/2),-45,1)
image_rotation03 = cv2.warpAffine(image_color,RotationMatrix,(cols,rows))
ShowImage([image_color, image_rotation01], 1, 2)
ShowImage([image_rotation02, image_rotation03], 1, 2)
In [9]:
# Shearing
from skimage import transform as tf
# Create Afine transform
afine_tf = tf.AffineTransform(shear=0.2)
image_shear01 = tf.warp(image_color, inverse_map=afine_tf)
afine_tf = tf.AffineTransform(shear=0.5)
image_shear02 = tf.warp(image_color, inverse_map=afine_tf)
afine_tf = tf.AffineTransform(shear=-0.2)
image_shear03 = tf.warp(image_color, inverse_map=afine_tf)
ShowImage([image_color, image_shear01, image_shear02, image_shear03], 2, 2)
2/ Kiếm 1 ảnh từ internet thực hiện các thao tác biến đổi sau bằng cách nhân thông qua ma trận biến đổi¶
- Scaling
- Rotation (30, 45, 60, 90, 120, 160, 180)
- Translation
- Shearing
In [10]:
# Read Image
image_color = imread("Sample08/monalisa.jpg")
# Convert Image into Gray
image_gray = cv2.cvtColor(image_color, cv2.COLOR_RGB2GRAY)
# Display Image
ShowImage([image_color, image_gray], 1, 2)
In [11]:
def Translate(image, shift_distance):
n,m = image.shape[:2]
dx = shift_distance[0]
dy = shift_distance[1]
scan = np.array([[1,0,dx],[0,1,dy]])
out_img = np.zeros(image.shape,dtype='u1')
for i in range(n):
for j in range(m):
origin_x = j
origin_y = i
origin_xy = np.array([origin_x,origin_y,1])
new_xy = np.dot(scan,origin_xy)
new_x = new_xy[0]
new_y = new_xy[1]
if 0 < new_x < m and 0 < new_y < n:
out_img[new_y,new_x] = image[i,j]
return out_img
In [12]:
shift_distance1 = (300,300)
shift_img1 = Translate(image_color, shift_distance1)
shift_distance2 = (-300,-300)
shift_img2 = Translate(image_color, shift_distance2)
shift_distance3 = (0,300)
shift_img3 = Translate(image_color, shift_distance3)
shift_distance4 = (300,300)
shift_img4 = Translate(image_color, shift_distance4)
ShowImage([shift_img1,shift_img2,shift_img3,shift_img4],2,2)
In [13]:
def resize(img, sx, sy):
height, width = img.shape[:2];
New_width = int(width * sx)
New_height = int(height * sy)
# xScale = New_width/(width-1)
# yScale = New_height/(height-1)
scale = np.array([[sy,0,0],[0,sx,0],[0,0,1]])
newImage = np.zeros((New_height, New_width, 3), dtype = "u1")
for i in range(height-1):
for j in range(width-1):
# newImage[i+1, j+1]= img[1 + int(i / yScale), 1 + int(j / xScale)]
original_xy = np.array([i,j,1])
new_xy = np.dot(scale,original_xy)
new_x = new_xy[1].astype(int)
new_y = new_xy[0].astype(int)
if 0 <= new_x < width and 0 <= new_y < height:
newImage[new_y,new_x] = img[i,j]
return newImage
In [14]:
# image_scale01 = cv2.resize(image_color,None,fx=0.5, fy=1, interpolation = cv2.INTER_CUBIC)
# image_scale02 = cv2.resize(image_color,None,fx=0.5, fy=0.5, interpolation = cv2.INTER_CUBIC)
# image_scale03 = cv2.resize(image_color,None,fx=1, fy=0.5, interpolation = cv2.INTER_CUBIC)
image_scale01 = resize(image_color,0.5,1)
image_scale02 = resize(image_color,0.5,0.5)
image_scale03 = resize(image_color,1,0.5)
print("\n image_color :")
InfoImage(image_color)
print("\n image_scale01 :")
InfoImage(image_scale01)
print("\n image_scale02 :")
InfoImage(image_scale02)
print("\n image_scale03 :")
InfoImage(image_scale03)
ShowImage([image_scale01, image_scale02, image_scale03], 1, 3)
In [15]:
#get rotation matrix
def getRMat(cx, cy, angle, scale):
a = scale*math.cos(angle*np.pi/180)
b = scale*(math.sin(angle*np.pi/180))
u = (1-a)*cx-b*cy
v = b*cx+(1-a)*cy
return np.array([[a,b,u], [-b,a,v]])
def rotation(image, angleInDegrees,scale=1):
height, width = image.shape[:2]
center_x, center_y = (width // 2, height // 2)
# rot = cv2.getRotationMatrix2D(img_centre, angleInDegrees, 1)
mat = getRMat(center_x,center_y, -int(angleInDegrees), scale)
cos = np.abs(mat[0,0])
sin = np.abs(mat[0,1])
new_width = int((height * sin) + (width * cos))
new_height = int((height * cos) + (width * sin))
mat[0, 2] += ((new_width / 2) - center_x)
mat[1, 2] += ((new_height / 2) - center_y)
outImg = cv2.warpAffine(image, mat, (new_width, new_height), flags=cv2.INTER_LINEAR)
return outImg
In [16]:
image_rotation01 = rotation(image_color,45)
image_rotation02 = rotation(image_color,75)
image_rotation03 = rotation(image_color,90)
InfoImage(image_rotation01)
InfoImage(image_rotation03)
ShowImage([image_rotation01,image_rotation02,image_rotation03],1,3)
In [17]:
def Shearing(image, sh, axis):
if axis == 1:
shear = np.array([[1,sh,0],[0,1,0],[0,0,1]])
elif axis == 0:
shear = np.array([[1,0,0],[sh,1,0],[0,0,1]])
height, width = image.shape[:2]
out_img = np.zeros((height,width,image.shape[2]), dtype=np.uint8)
for i in range(height):
for j in range(width):
origin_xy = [i,j,1]
new_xy = np.dot(shear,origin_xy)
new_x = new_xy[1].astype(int)
new_y = new_xy[0].astype(int)
if 0 <= new_x < width and 0 <= new_y < height:
out_img[i,j] = image[new_y,new_x]
return out_img
In [18]:
shearing1 = Shearing(image_color,0.3,1)
shearing2 = Shearing(image_color,-0.3,1)
shearing3 = Shearing(image_color,0.3,0)
shearing4 = Shearing(image_color,-0.3,0)
ShowImage([shearing1,shearing2,shearing3,shearing4],2,2)
3/ Cũng thực hiện nhân ma trận. Tạo các ma trận tổng hợp sau:¶
- Scaling + Rotation
- Rotation + Shearing
- Scaling + Rotation + Shearing
In [19]:
# scaling + rotation
new_image1 = rotation(resize(image_color,1.5,1.5),45)
new_image2 = rotation(resize(image_color,2,1),90)
ShowImage([new_image1,new_image2],1,2)
In [20]:
img1 = resize(rotation(image_color,45),1.3,0.7)
img2 = resize(rotation(image_color,75),0.5,0.9)
ShowImage([img1,img2],1,2)
In [21]:
# rotation + shearing
new_image1 = rotation(Shearing(image_color,0.3,0),45)
new_image2 = rotation(Shearing(image_color,-0.3,1),50)
ShowImage([new_image1,new_image2],1,2)
In [22]:
new_image1 = Shearing(rotation(image_color,45),-0.3,0)
new_image2 = Shearing(rotation(image_color,75),-0.3,1)
ShowImage([new_image1,new_image2],1,2)
In [23]:
# scaling + rotation + shearing
new_image1 = rotation(resize(Shearing(image_color,0.5,1),1.5,1.5),75)
new_image2 = Shearing(rotation(resize(image_color,0.7,0.5),45),-0.5,1)
new_image3 = rotation(Shearing(resize(image_color,0.5,1.2),0.5,1),75)
ShowImage([new_image1,new_image2,new_image3],1,3)
4/ Chụp một tấm ảnh bất kỳ từ điện thoại của bạn có chứa vật thể dạng hình chữ nhật hay hình vuông (ví dụ bức tranh, điện thoại, màn hình máy tính v.v…) (Lưu ý chụp nghiêng, xéo và không được chụp thẳng)¶
- Xác định 4 điểm góc của đối tượng bằng tay và vẽ chúng lên hình
- Dùng biến đổi Affine hoặc Perspective để canh chỉnh chúng thẳng
In [24]:
# Read Image
image_color = imread("Sample08/My_ID.png")
image_color = ResizeImage(image_color, 500, 800)
# Convert Image into Gray
image_gray = cv2.cvtColor(image_color, cv2.COLOR_RGB2GRAY)
# Display Image
InfoImage(image_color)
ShowImage([image_color, image_gray], 1, 2)
In [25]:
rows,cols,ch = image_color.shape
PointList_From = [[170,180],[370,460],[0,300],[160,540]]
PointList_To = [[0,0],[300,0],[0,300],[300,300]]
XList = []
YList = []
for pt in PointList_From:
XList.append(pt[0])
YList.append(pt[1])
plt.figure(figsize=(10,10))
plt.imshow(image_color)
plt.scatter(XList, YList, c='y', s=200, marker='^')
plt.plot(XList,YList,'r', linewidth=5, markersize=12)
plt.show()
pts1 = np.float32(PointList_From)
pts2 = np.float32(PointList_To)
MappingMatrix = cv2.getPerspectiveTransform(pts1,pts2)
image_Perspective = cv2.warpPerspective(image_color,MappingMatrix,(300,300))
ShowImage([image_color, image_Perspective], 1, 2)
XList = []
YList = []
for pt in PointList_To:
XList.append(pt[0])
YList.append(pt[1])
plt.figure(figsize=(10,10))
plt.imshow(image_Perspective)
plt.scatter(XList, YList, c='y', s=200, marker='^')
plt.plot(XList,YList,'r', linewidth=5, markersize=12)
plt.show()
